neurovox
NeuroVox detects Parkinson's disease in its earliest stages using vocal biomarkers and machine learning. by analyzing acoustic features of speech — jitter, shimmer, harmonic-to-noise ratio, fundamental frequency — the system achieves 94.87% test accuracy and 100% recall, supporting physicians with data-driven diagnostics through a user-friendly interface.
the diagnostic challenge
Parkinson's is a progressive neurodegenerative disorder that impacts motor function, speech, and quality of life. in its early stages, it's notoriously hard to detect. minor vocal changes — reduced pitch variation, subtle tremors — often go unnoticed by the human ear and standard clinical assessments. by the time a diagnosis is made, significant neurodegeneration has often already occurred. earlier detection means earlier intervention.
machine learning approach
NeuroVox trains classification models on a dataset of voice recordings, extracting acoustic features that correlate with the vocal impairments characteristic of early Parkinson's. the full preprocessing pipeline includes duplicate removal, outlier detection, MinMax normalization, SMOTE balancing for class imbalance, and cross-validation. six models were evaluated: Logistic Regression, SVM, Decision Tree, Random Forest, KNN, and Neural Networks.
performance results
Random Forest and KNN each achieved 94.87% test accuracy with 100% recall and F1-scores of 96.97. 100% recall means every actual Parkinson's case in the test set was detected — critical in any medical application where false negatives carry serious consequences. the system was awarded gold at the IJAS State Competition.
clinical integration
NeuroVox is deployed as a user-facing interface where patients or clinicians can upload voice samples and receive real-time diagnostic feedback. the design prioritizes accessibility for integration into telemedicine platforms, making early-stage diagnostics available in regions without access to neurologists. it supports clinical judgment rather than replacing it — a powerful second opinion grounded in data.